Together AI

The Together AI provider uses provider key together and the HPD-Agent.Providers.Together package. ModelName is a Together chat model id for chat, or a Together embedding model id for embeddings.

The provider package is currently net10.0 only because the underlying Together SDK targets net10.0.

Set an API key:

export TOGETHER_API_KEY="..."

The default endpoint is https://api.together.ai/v1. Pass endpoint to WithTogether(...) / WithTogetherEmbeddings(...), or set Endpoint in ChatClientConfig, when you need a compatible proxy or custom Together base URL.

Use fluent chat setup first:

using HPD.Agent;
using HPD.Agent.Providers.Together;

var agent = await new AgentBuilder()
    .WithTogether(model: "meta-llama/Llama-3.3-70B-Instruct-Turbo")
    .BuildAsync();

var result = await agent.RunAsync("Write one sentence about Together AI setup.");
Console.WriteLine(result.Text);

The equivalent Clients.Chat shape is:

{
  "Clients": {
    "Chat": {
      "ProviderKey": "together",
      "ModelName": "meta-llama/Llama-3.3-70B-Instruct-Turbo"
    }
  }
}

Embeddings

Together also registers the embeddings provider family under the same together key:

using HPD.Agent;
using HPD.Agent.Providers.Together;

var agent = await new AgentBuilder()
    .WithTogetherEmbeddings(model: "BAAI/bge-large-en-v1.5")
    .BuildAsync();

The equivalent family config is:

{
  "Clients": {
    "Embeddings": {
      "ProviderKey": "together",
      "ModelName": "BAAI/bge-large-en-v1.5"
    }
  }
}

Runtime Chat Options

Use ChatClientConfig for shared model-call behavior such as temperature, top-p, top-k, max output tokens, stop sequences, response format, tools, tool mode, seed, and generic reasoning options. Put agent-level defaults in AgentConfig.Clients.Chat; use AgentRunConfig.Clients.Chat for complete per-run replacements.

The Together SDK adapter maps the common Microsoft.Extensions.AI chat options to the Together request. HPD.Agent also maps generic frequency penalty, presence penalty, and Reasoning into Together's raw request fields so HPDOS/model metadata can keep using provider-neutral chat config.

Use TogetherChatRequestOptions only for Together-specific request fields:

using HPD.Agent;
using HPD.Agent.Providers.Together;

var agent = await new AgentBuilder()
    .WithTogether(model: "deepseek-ai/DeepSeek-R1")

    .WithTogetherChatRequestOptions(options =>
    {
        options.ContextLengthExceededBehavior = TogetherContextLengthExceededBehavior.Truncate;
        options.RepetitionPenalty = 1.1;
        options.Logprobs = 5;
        options.MinP = 0.05f;
        options.SafetyModel = "safety_model_name";
    })
    .BuildAsync();

The typed Together request options are serializable and can also be applied to a run/session chat config:

var runConfig = new AgentRunConfig
{
    Clients = new AgentClientsConfig
    {
        Chat = new ChatClientConfig
        {
            ProviderKey = "together",
            ModelName = "deepseek-ai/DeepSeek-R1",
            Reasoning = new ReasoningOptions
            {
                Effort = ReasoningEffort.Medium
            }
        }
    }
};

runConfig.Clients.Chat!.UseTogetherChatRequestOptions(new TogetherChatRequestOptions
{
    ContextLengthExceededBehavior = TogetherContextLengthExceededBehavior.Error,
    ReasoningEnabled = true
});

Available Together request options:

Option	Request field	Use
ContextLengthExceededBehavior	context_length_exceeded_behavior	Chooses error or truncate when requested tokens exceed model context.
RepetitionPenalty	repetition_penalty	Penalizes repeated sequences.
Logprobs	logprobs	Returns top token log probabilities for each generation step.
Echo	echo	Includes the prompt in the response.
N	n	Requests multiple completions for each prompt.
MinP	min_p	Uses minimum probability sampling as an alternative to top-p/top-k.
LogitBias	logit_bias	Biases specific token ids.
Compliance	compliance	Passes a Together compliance mode value.
ChatTemplateKwargs	chat_template_kwargs	Sends additional model-engine chat template settings.
SafetyModel	safety_model	Selects a Together moderation model.
ReasoningEnabled	reasoning.enabled	Overrides generic reasoning enable/disable behavior.

TogetherProviderConfig exposes Together-specific embedding defaults, such as EmbeddingModelId, for the embeddings family.

For embeddings, WithTogetherEmbeddings(...) stores the embedding model in the embeddings family config.

Caveats

Together chat and embeddings are available only when the HPD-Agent.Providers.Together package is referenced by a net10.0 application.

Together chat supports token streaming, function tools, reasoning content, and JSON object responses through the Microsoft.Extensions.AI adapter. Vision, image generation, rerank, and audio families are not registered by this provider.

Runtime validation checks model, endpoint, and embedding model options. Runtime chat options are applied through Clients.Chat and AgentRunConfig.Clients.Chat. Credentials are resolved lazily when a run first acquires this provider client; a missing credential fails that run, not BuildAsync().

Related Reading

• Provider Setup Overview
• Provider Families
• Provider Keys And Environment Variables